Full text of DOI:10.1053/ajkd.2002.30549

Long-Term Glycemic Control Measurements in Diabetic Patients
Receiving Hemodialysis
Melanie S. Joy, PharmD, William T. Cefalu, MD, Susan L. Hogan, PhD,
and Patrick H. Nachman, MD
● Cardiovascular morbidity is increased in patients with diabetes mellitus and there is a great prevalence of
diabetes and cardiovascular disease among patients with end-stage renal disease (ESRD). Control of glycemia can
decrease cardiovascular and end-organ damage. Because the validity of glycemic control tests have not been
rigorously studied in patients with ESRD, we evaluated the value of various measures in these patients. The overall
clinical goal was to investigate whether hemoglobin A1C (A1C) accurately reflects actual glycemic control as
compared with other measures in light of the importance of attaining appropriately controlled blood glucose (BG).
The commonly used tests of total glycated hemoglobin (GHb) and A1C may be unreliable in patients with ESRD
because of the presence of anemia, shortened red blood cell (RBC) survival, and assay interferences from uremia.
The primary aim of this study was to assess the relationship of capillary BG measurements to A1C, GHb, total
glycated plasma proteins (GPP), and fructosamine (Fr) in diabetic patients receiving hemodialysis. Twenty-three
patients were instructed to obtain BG evaluations twice daily for 7 days by using the Elite glucometer (Bayer
Corporation, Elkhart, IN). These determinations included 6 fasting, 6 preprandial, and 3 separate 2-hour postpran-
dial levels. Blood was obtained on day 7 for measurement of A1C, GHb, GPP, and Fr. A1C was analyzed by an
immunoassay, GPP and GHb were assayed by affinity high-performance liquid chromatography (HPLC), and Fr by
automated nitroblue colorimetric assay. Scatter plots were generated by plotting the average BG versus A1C, GHb,
GPP, or Fr. Linear regression was performed for each plot showing the following relationships: A1C ؍ 0.0174 (BG) ؉
4.76 (r ؍ 0.58; P < 0.05): GHb ؍ 0.0371 (BG) ؉ 3.57 (r ؍ 0.584; P < 0.05): GPP ؍ 0.0083 (BG) ؉ 26.13 (r ؍ 0.065; P ؍
0.77): Fr ؍ 0.6865 (BG) ؉ 250 (r ؍ 0.345; P ؍ 0.11). Despite anemia and shortened RBC lifespan in patients with
ESRD, A1C in the range of 6% to 7% estimates glycemic control similarly to patients without severe renal
impairment. A1C values above 7.5% may overestimate hyperglycemia in patients with ESRD. Thus, diabetic patients
receiving hemodialysis may have long-term BG that are more properly controlled than previously determined,
reducing their risks of the macro- and microvascular complications of diabetes mellitus.
© 2002 by the National Kidney Foundation, Inc.
INDEX WORDS: Hemoglobin A1C; glycemic control; hemodialysis (HD); diabetes.
IABETES MELLITUS represents 37% of
all prevalent cases of end-stage renal dis-
glycated hemoglobin (hemoglobin A1C [A1C]).
The same accepted normal ranges for A1C are
applied to all diabetic patients, regardless of their
renal function. The good correlation between
A1C and BG in non-ESRD type 1 diabetic pa-
tients has been documented in the Diabetes Con-
trol and Complications Trial (DCCT).14,15 An
D
ease (ESRD) patients receiving dialysis.1 Al-
though a higher frequency of ESRD is seen
in patients with type 1 diabetes (40% develop
ESRD), the actual prevalence is greater in type
2 diabetes because of the greater prevalence of
type 2 diabetes in the population.2 Other long-
term complications resulting from diabetes
include retinopathy, neuropathy, cardiovascular
disease, and their sequelae. Control of hypergly-
cemia is known to decrease short-term morbid-
ity, small-vessel disease, and cardiovascular
risks.3-5 In addition, control of other concomitant
factors (hyperlipidemia, tobacco abuse, blood
pressure, and dietary protein intake), as well as
therapy with angiotensin-converting enzyme in-
hibitors or angiotensin-receptor blockers, have
slowed the progression of complications in pa-
tients with and without diabetes mellitus.6-13
Clinically, measurement of glycemic control
in diabetic hemodialysis patients has been deter-
mined in a similar manner as for patients without
ESRD; by monitoring blood glucose (BG) and
From the Division of Nephrology and Hypertension, Uni-
versity of North Carolina, School of Medicine, Chapel Hill,
NC; and the Endocrine, Diabetes, and Metabolism Unit,
University of Vermont, College of Medicine, Burlington, VT.
Received June 15, 2001; accepted in revised form August
31, 2001.
Supported in part by a grant from the North Carolina
National Kidney Foundation.
Presented in part as a poster presentation at the 31st
annual meeting of the American Society of Nephrology,
October 25-28, 1998, Philadelphia, PA.
Address reprint requests to Melanie S. Joy, PharmD,
Clinical Assistant Professor, University of North Carolina,
Schools of Medicine and Pharmacy, Division of Nephrology
and Hypertension, CB #7155, 348 MacNider Building, Chapel
Hill, NC 27599-7155. E-mail: Melanie_Joy@med.unc.edu
© 2002 by the National Kidney Foundation, Inc.
0272-6386/02/3902-0008$35.00/0
doi:10.1053/ajkd.2002.30549
American Journal of Kidney Diseases, Vol 39, No 2 (February), 2002: pp 297-307
297

298
JOY ET AL
Table 1. Contributing Factors Complicating
Evaluation of Long-Term Glycemic Control
average BG concentration can be estimated based
on the glycated hemoglobin and vice versa.14,15
Because the DCCT did not include patients with
ESRD and very few patients had progressive
renal disease (defined by increased proteinuria),
this assumption may not be valid in this patient
population.16 There are reports in the literature of
falsely increased values of A1C and total gly-
cated hemoglobin (GHb) in patients with renal
failure caused by carbamylated hemoglobin inter-
fering with the assay.17 Other factors known to
increase A1C measurements include: increased
hemoglobin F, hypertriglyceridemia, hyperbiliru-
binemia, opiate and alcohol addiction, lead poi-
soning, uremia, and high-dose aspirin therapy.17,18
Falsely decreased A1C measurements have been
documented owing to decreased red blood cell
(RBC) mass, decreased RBC survival, iron-
deficiency anemia, phlebotomy, changes in iron
or erythropoietin administration or hemolytic
anemia (increases in erythrocyte pool), and he-
modilution caused by blood transfusions.17-23 Sev-
eral of these factors that can effect A1C measure-
ments are found in patients with ESRD. These
confounding factors suggest that the correlations
of A1C and BG concentrations may not be as
predictable and accurate in patients with ESRD
and, thus, may be a misleading assessment of the
true degree of glycemic control. For this reason,
there has been an interest in other markers of
glycemic control such as glycated plasma pro-
teins, that are independent of RBC survival.
Unfortunately, the use of glycated plasma pro-
teins may also be limited by various confound-
ers. Falsely decreased values have been shown in
the presence of low protein concentrations.17
Falsely increased values have been measured in
the presence of lipemia, hemolysis, high biliru-
bin and uric acid concentrations, uremia, and in
patients receiving high doses of aspirin therapy.17
The potential confounders to the use of glycated
hemoglobin and glycated plasma proteins are
summarized in Table 1.
Falsely
Falsely
Increased Values
Decreased Values
Glycated
hemoglobin
Carbamylated
hemoglobin
Reduced RBC
lifespan
Uremic acidosis
Labile glycated
compounds
RBC transfusions
Vitamins C and E
Hemoglobin F
Pregnancy
Hypertriglyceridemia Iron deficiency
Opiate addiction
Lead poisoning
Alcoholism
Phlebotomy
Hemolytic
anemia
High doses of
aspirin
Glycated
plasma
proteins
Lipemia
Hyperbilirubinemia
Hemolysis
Low serum
protein
concentrations
Increased uric acid
Uremia
High doses of
aspirin
chronic hemodialysis; (2) characterize RBC sur-
vival by erythrokinetic modeling to determine if
this factor accounts for changes in correlation
between A1C and BG in patients with ESRD; and
(3) develop a regression model to relate the BG
values to corresponding A1C values in the ESRD
patient population receiving hemodialysis and
compare this with the regression relationship
published for diabetic patients with normal renal
function based on the results of the DCCT.
METHODS
Diabetic patients and their controls receiving hemodialy-
sis at the University of North Carolina Hospitals dialysis
units were identified. Patient inclusion criteria consisted of:
(1) age greater than 18 years, (2) receiving hemodialysis 3
times weekly, and (3) diagnosis of type 1 or type 2 diabetes
mellitus. The exclusion criteria were: (1) receipt of a blood
transfusion within the previous 3-month time period, (2)
inability or unwillingness to perform the required BG mea-
surements, (3) diagnosis of an immune or hereditary hemo-
lytic anemia, (4) treatment with drugs known to interfere
with the assays for glycemic control measures (eg, high-
dose vitamin C, vitamin E, or aspirin, and opiate or alcohol
addiction), and (5) serum albumin concentrations less than
3.0 g/dL. Patients were dialyzed in accordance with the
recommendations presented in the Dialysis Outcomes Qual-
ity Initiative (DOQI) practice guidelines for hemodialysis
adequacy.24 Twenty-four patients signed a consent form to
participate (23 were evaluable). Nine patients participated
on a second occasion to determine intraindividual variations
This study was designed to evaluate 4 meth-
ods of measuring BG control in chronic hemodi-
alysis patients with diabetes mellitus. The spe-
cific aims of the study were to: (1) measure and
evaluate the correlation betweenA1C, GHb, fruc-
tosamine (Fr), and total glycated plasma proteins
(GPP) as indicators of glucose control when
compared with BG testing in diabetic patients on

GLYCEMIC CONTROL MEASURES IN HEMODIALYSIS
299
across time. In addition, 8 nondiabetic dialysis patients were
selected to serve as controls. All diabetic patients were
instructed to use the Glucometer Elite BG meter (Bayer
Corporation, Elkhart, IN) to obtain twice-daily BG measure-
ments for 7 consecutive days. The product literature reported
precision results from diabetic patients with mean high and
low BG values of 260 mg/dL and 58.3 mg/dL, respectively.
The within and overall coefficient of variations for the high
and low were 3.1% and 4.1% (within) and 5.1% and 5.7%
(overall), respectively.
CA) for the purpose of this study. This model does not adjust
for transfusions, intravenous iron administration, dialysis
adequacy, hyperparathyroidism, and other factors known to
affect erythropoietin sensitivity. We retrospectively entered
erythropoietin doses and measured hematocrit levels into the
EPOCALC program for a period of 3 to 6 months before our
assessment periods for A1C and GHb. Pharmacodynamic
modeling provided information regarding RBC lifespan and
individual sensitivity to erythropoietin for 18 of our patients.
The RBC lifespan of patients not receiving erythropoietin ␣
therapy could not be determined because the EPOCALC
program is based on average erythropoietin sensitivity and
RBC lifespan among ESRD patients receiving erythropoi-
etin.
The instruction for the glucometer was repeated several
times by the study personnel. The patients had to show their
competence in performing these tests before actually partici-
pating in the study. The BG measurements consisted of a
predefined variation of 6 fasting, 6 preprandial, and 3 post-
prandial assessments (2 hour), obtained over a week. The
DCCT used a 24-hour, 7-point, BG profile method in the
intensive therapy arm. The 7-point BG profile was per-
formed quarterly, whereas the A1C was tested monthly.14
Because of the day-to-day variation in caloric and carbohy-
drate intake, and level of physical stress imposed by the 3
times per week hemodialysis regimen, we felt that a 7-point
BG profile over a single day would be less representative of
glycemic control than a set of measures obtained over a
week. The inter- and intradialysis treatment periods over the
course of a week should more adequately reflect the fluctua-
tions in insulin requirements seen in dialysis patients with
diabetes. Therefore, we modified the BG sampling scheme
to a 14-point scheme over a 7-day period. Patients would be
less likely to alter their normal diet and subsequent BG
levels during this longer time period, thus making these
measurements more reflective of each patient’s usual glyce-
mic control. This scheme did allow for assessment of blood
sugars preprandial, at bedtime, and postprandial (breakfast,
lunch, and dinner), as defined.
A data collection form was provided to each patient to
record the test results. The glucometer stored the results of
these tests in memory so that confirmation of the patient-
reported data could be assessed for accuracy. When a discrep-
ancy resulted between the written test results and those
recorded in the glucometer’s memory, the values stored in
memory were used to calculate average BG. On the final day
of testing (day 7 dialysis session), blood (15 mL) was
collected for assessment of measures of glycemic control
before infusion of heparin. The tubes were centrifuged at
2,000 ϫ gravity, and blood, plasma, and serum were col-
lected and assayed:A1C (blood), GHb (blood), GPP (plasma),
and Fr (serum). The control patients had their blood drawn
for a random BG, A1C, GHb, GPP, and Fr predialysis. The
charts of all patients were reviewed for demographic, medi-
cal, medication, and additional laboratory information. This
study was reviewed and approved by the institutional review
board at the University of North Carolina and conducted
according to the Declaration of Helsinki.
Assays
Hemoglobin A1C concentrations were assessed by an
immunoassay (Tina-Quant; Boehringer Mannheim, Indianap-
olis, IN), by using a BM/Hitachi 717 analyzer (Roche
Diagnostics Corp, Indianapolis, IN). This method was with-
out significant (Ϫ0.2% to 0.10%) A1C changes caused by
bilirubin (Յ50 mg/dL), lipemia (Յ800 mg/dL), ascorbic
acid (Յ50 mg/dL), and rheumatoid factors (Յ750 IU/mL).
The test was specific without cross-reactivity to carbamy-
lated hemoglobin, glycated albumin, labile A1C, or other
hemoglobin A subfractions. The technical limits of the assay
were 0.2 g/dL A1C up to the value of the highest calibrator
(2.5 g/dL), and hemoglobin concentrations of 6 to 24 g/dL.
Calculation of A1C was made by using the following for-
mula from the product literature: %A1C ϭ (A1C (g/dl)/Hgb
(g/dl) * m) ϩ b, where Hgb ϭ hemoglobin, m ϭ slope, and
b ϭ intercept. Total coefficient of variation was less than
5.0% for A1C. The bias in this method when compared with
the DCCT trial high performance liquid chromatography
(HPLC) method was Ϫ0.09 (for an A1C of 6%) and Ϫ0.1
(for an A1C of 8%). Previously reported data showed excel-
lent correlation and similar line of identity between the
DCCT HPLC and Tina-Quant methods for determination of
A1C values of 15% or less.26,27
GHb represents all subfractions of hemoglobin, including
A1C. The GPP measures all GPP. Both GHb and GPP were
assayed by automated affinity HPLC on a Primus CLC-330
HPLC (Primus Corp., Kansas City, MO) as previously
described.28 Serum Fr, which measures the ability of ket-
amines to act as reducing agents, was assayed on a Cobas
Mira Chemistry analyzer by using Roche reagents (Roche
Diagnostic Systems, Nutely, NJ) as previously described.29,30
Statistical Analysis
The demographics and baseline measurements (body mass
index [BMI], albumin) between diabetic and nondiabetic
patients were compared by using Fisher’s Exact tests for
categoric measures and Wilcoxon rank tests for continuous
measures. The relationship between BG and each of the 4
measures of glycemic control (A1C, GHb, GPP, and Fr) were
observed with graphic plots. Linear regression was used to
model the relationship of BG with each of the 4 measures
and Pearson’s correlation coefficients were computed for
each association. Multivariate regression analysis was used
to control for factors such as race, sex, BMI, and RBC
RBC Lifespan Method
The method of Uehlinger et al25 was used to estimate RBC
lifespans in our dialysis patients. A computerized model
of this approach that incorporates administered dosages
of erythropoietin ␣ and measured hematocrit levels
(EPOCALC) was supplied by Amgen, Inc. (Thousand Oaks,

300
JOY ET AL
Table 3. Glycemic Control Laboratory Results
lifespan. Only measures associated with BG (race) or mea-
sures found to be different between diabetic and nondiabetic
patients (BMI) are reported in multivariate models. Other
variables (sex, age, and albumin) did not add to the predic-
tion of the models or confound the relationship between BG
and any of the other measures. RBC lifespan was also
controlled for in the subset of diabetic patients because there
is a postulated relationship between survival of RBCs and
subsequent results of glycemic control measures using hemo-
globin subfractions. Model-adjusted correlation coefficients
were reported for multivariate regression models for general
comparison with the univariate correlation coefficients. Be-
cause some patients (n ϭ 9) had a repeat testing period 6
months later, univariate regression analysis was used with
the increased number of episodes (n ϭ 32) to compare with
the single-episode measurements. The regression lines evalu-
ating BG and A1C relationships from this study and the
DCCT trial were graphed. The 95% confidence intervals for
the slope and intercept from our study were computed.
These values were not available for the DCCT trial.
Diabetes
(n ϭ 23)
Nondiabetic Patients
P
Value
(n ϭ 8)
BG (mg/dL)
A1C (%)
GHb (%)
GPP (%)
160.2 Ϯ 43.2
7.5 Ϯ 1.3
9.5 Ϯ 2.7
92.4 Ϯ 23.7
5.4 Ϯ 0.2
5.5 Ϯ 0.4
20.6 Ϯ 3.8
286.4 Ϯ 59.6
0.0012
0.0004
0.0004
0.0071
0.049
27.5 Ϯ 5.5
Fr (mcmoles/L) 360 Ϯ 85.9
(range 6.7%–16.5%). The nondiabetic dialysis
patients exhibited A1C and GHb of 5.4 Ϯ 0.2 and
5.5 Ϯ 0.4, respectively, which were within the
normal ranges for nondiabetic patients who are
not dialysis dependent. The GPP averaged 27.5%
(range, 21.1%–27.7%) in diabetic dialysis pa-
tients. The Fr assay averaged 360 mcmoles/L
(range, 292–586 mcmoles/L). Although a normal
reference range has not been published for GPP
or Fr in nondiabetic dialysis patients for compari-
son, our values were 20.6% Ϯ 3.8% and 286.4 Ϯ
59.6 mcmoles/L, respectively. Table 4 provides a
comparison between the measures of glycemic
control in nondiabetic patients who were hemodi-
alysis dependent (from our study) versus litera-
ture values for patients without renal dis-
ease.28,31,32 The ranges of these values for patients
receiving hemodialysis overlapped with the
ranges for patients without renal disease for all
measures except Fr and GPP. For both Fr and
GPP, the upper limit of the range was greater in
hemodialysis patients than those not receiving
hemodialysis. For all of these measures, the
lower limits of the ranges were higher in patients
receiving hemodialysis than in those without
renal failure.
RESULTS
The demographics of the diabetic and control
hemodialysis patients are presented in Table 2.
The age, race, sex, and albumin concentrations
were similar between groups. The diabetic pa-
tients were nearly equally divided between sex
and races. The BMI was significantly higher in
patients with diabetes (P ϭ 0.04). All study
patients received hemodialysis for 3.5 to 4.5
hours, 3 times weekly. The blood and dialysate
flow rates ranged from 250 to 450 mL/min and
450 to 600 mL/min, respectively. The delivered
dialysis doses (KT/V) were similar for nondia-
betic and diabetic patients (P ϭ 0.097).
The values for glycemic control measures are
presented in Table 3. Based on the recommended
target value of A1C (Ͻ7.0%), our diabetic pa-
tients on hemodialysis had, on average, reason-
ably good glycemic control (mean ϭ 7.5%, range
of A1C, 5.8%–11.2%). The GHb averaged 9.6%
Average BG was plotted against A1C, GHb,
GPP, and Fr to determine the correlations and
Table 4. Reported Reference Ranges for Glycemia in
Patients Without Diabetes for Assay Methods
Used in Current Study
Table 2. Patient and Control Demographics
Diabetes
Nondiabetic Patients
P
(n ϭ 23)
(n ϭ 8)
Value
Nonhemodialysis
Hemodialysis*
Sex
(men/women) 13/10
7/1
5/3
57 Ϯ 17
24.8 Ϯ 4.1
3.8 Ϯ 0.4
1.56 Ϯ 0.2
0.203
0.433
0.727
0.041
0.928
0.097
Hemoglobin
A1C
Total GHb
Fr
Race (W/AA)
Age (yr)
BMI
10/13
56 Ϯ 11
30.1 Ϯ 5.9
4.0%–5.5%31
4.0%–8.0%28
5.2%–5.6%
4.7%–6.1%
165–303 mcmoles/L32 205–370 mcmoles/L
Albumin (g/dL) 3.8 Ϯ 0.3
Kt/V 1.41 Ϯ 0.2
Total GPP
13.4%–25.0%28
17.7%–26.7%
*The ranges for patients on hemodialysis are derived
from the current study.
Abbreviations: W, White; AA, African American.

GLYCEMIC CONTROL MEASURES IN HEMODIALYSIS
301
Fig 3. Linear regression analysis of average BG
and total glycosylated plasma proteins showed the
weakest correlation of all the measures: GPP ؍ 0.0083
(BG) ؉ 26.13 (r ؍ 0.065; P ؍ 0.77).
Fig 1. Linear regression analysis of average BG
and hemoglobin A1C showed similar correlation to the
total glycosylated hemoglobin relationship (Fig 2):
A1C ؍ 0.0174 (BG) ؉ 4.76 (r ؍ 0.579; P < 0.05).
When compared with Fr and GPP, the graphs
that plotted A1C or GHb and average BG showed
the best correlations. The relationship between
BG and GPP was the least strongly correlated.
This relationship is defined by the following
regression equation and Fig 3:
equation of the line representing the relation-
ships. A linear relationship between BG and each
of the 4 measures of glycemic control (ie, A1C,
GHb, Fr, and GPP) were determined by graphic
analysis.
The regression equations defined in Figs 1 and
2 are as follows:
GPP ϭ 0.0083 (BG)
A1C ϭ 0.0174 (BG)
ϩ 26.13 (r ϭ 0.065; P ϭ 0.77)
ϩ 4.76 (r ϭ 0.579; P Ͻ 0.05)
Fructosamine was poorly correlated with BG.
The following regression equation and Fig 4
defines this relationship:
GHb ϭ 0.0371 (BG)
ϩ 3.57 (r ϭ 0.584; P Ͻ 0.05)
Fig 4. Linear regression analysis of average blood
glucose and fructosamine showed a poor correlation
when compared with total glycosylated hemoglobin
and hemoglobin A1C: Fr ؍ 0.687 (BG) ؉ 250 (r ؍ 0.345;
P ؍ 0.11).
Fig 2. Linear regression analysis of average BG
and total glycosylated hemoglobin showed similar cor-
relation to the hemoglobin A1C relationship (Fig 1):
GHb ؍ 0.0371 (BG) ؉ 3.57 (r ؍ 0.584; P < 0.05).

302
JOY ET AL
nificances were seen with these models, the unad-
justed equations are reported for all associations.
Erythrokinetic modeling for the diabetic pa-
tients receiving erythropoietin therapy showed
an average RBC lifespan of 53.1 Ϯ 18.6 days.
This value is below the average RBC lifespan
(68 days) determined in hemodialysis patients by
Uehlinger et al.25 A plot of RBC lifespan versus
A1C showed a poor relationship and correlation
(A1C ϭ .010 [RBC lifespan] ϩ 6.810, r ϭ .147,
P ϭ .56) (Fig 5). This model was not improved
significantly with the addition of the repeated
measurements that were performed in a small
group of patients. Analysis of variance testing to
control for BG in the model of RBC lifespan and
A1C did not improve the relationship.
When compared with the DCCT trial of diabetic
patients without severely compromised renal func-
tion, there was a trend for A1C values to be associ-
ated with lower BG in diabetic dialysis patients at
A1C values greater than 7% (Fig 6). On the other
hand, at lower A1C values (Ͻ6%), higher BG
values were observed in hemodialysis patients than
in DCCT patients. The maximum limit of the 95%
confidence interval for the mean regression line
estimated for the diabetic dialysis patients ap-
proaches the regression line of the DCCT trial (Fig
6). However, no measures of variation were avail-
able from the DCCT data (personal communica-
Fig 5. A poor correlation between RBC lifespan and
hemoglobin A1C was shown: A1C ؍ 0.010 (RBC life-
span) ؉ 6.810 (r ؍ 0.147, P ؍ 0.56).
Fr ϭ 0.687 (BG)
ϩ 250 (r ϭ 0.345; P ϭ 0.11)
Because of the poor correlation between Fr
and BG, we sought to determine the role, if any,
albumin concentration may have because the
literature suggests a relationship between plasma
albumin and Fr.32 The relationship between albu-
min and Fr was poorly and nonsignificantly
correlated. We calculated the Fr concentrations
adjusted for the degree of hypoalbuminemia as
suggested by Howey et al.33 A graph of BG
versus adjusted Fr showed a similar regression
equation and correlation coefficient as for the
unadjusted Fr graph (data not shown). Based on
the similarity between groups, this adjustment
did not appear to be necessary for our hemodialy-
sis patients with albumin concentrations of 3.0
mg/dL or greater. All of the earlier-mentioned
regression models were redrawn with the addi-
tion of the repeated studies with 9 patients (n ϭ
32). The equations representing these models
were similar to the single-measurement models,
except for the correlation between BG and Fr,
which became significant with the additional
measurements [Fr ϭ 1.092 (BG) ϩ 201 (r ϭ
0.481; P ϭ .005)].
Fig 6. The upper limit of the 95% confidence inter-
val for the diabetic dialysis patients overlaps with the
regression line of the DCCT trial. For the DCCT, the
published regression equation was: BG ؍ 30.9 (A1C) ؊
60.2. The regression equation for our data was BG ؍
19.28 (A1C) ؉ 14.85. The 95% confidence intervals for
our data were computed for slope (7.7, 30.9) and inter-
cept (؊74.0, 103.7). Variation estimates were not avail-
able from the DCCT investigators.
Multivariate models controlling for race, BMI,
and RBC lifespan were evaluated for each equa-
tion (A1C, GHb, GPP, and Fr) with BG. Because
no appreciable differences in parameter esti-
mates, correlation coefficients, and statistical sig-

GLYCEMIC CONTROL MEASURES IN HEMODIALYSIS
303
tion), preventing statistical comparisons of the 2
lines. These results suggest that A1C may underes-
timate the long-term level of glycemic control in
diabetic patients requiring hemodialysis, especially
at A1C levels above 7.5%.
incompletely evaluated. In question is not only
whether A1C correlates with glycemic control in
patients with ESRD, but whether A1C measures
correspond to the same average serum glucose
concentrations in patients with ESRD and pa-
tients with normal renal function. It is important
to determine if currently used measures of GHb
over- or underestimate glycemic control in pa-
tients with ESRD. Of particular concern would
be if A1C overestimates glycemic control in
patients with ESRD (ie, measured A1C is lower
than expected for the same glucose concentra-
tions in patients without renal insufficiency).
This would in effect expose diabetic patients
with ESRD to an increased risk for long-term
small- and large-vessel disease associated with
poor glycemic control, such as retinopathy and
neuropathy. In addition, because 50% of patients
commencing renal replacement therapy suffer
from cardiovascular disease, the impact of inad-
equately controlled diabetes on cardiovascular
and cerebrovascular risks may be significant in
this population.36
In this study, we aimed to evaluate the correla-
tion and accuracy of common clinically used
measures of glycemic control (A1C and GHb) in
chronic hemodialysis patients with diabetes mel-
litus. Because of the potential for limitations on
the accuracy of measures of glycemic control
that are based on hemoglobin, we evaluated GPP
and Fr as possible alternative measures, perhaps
better suited to the ESRD population. The ratio-
nale for using glycated serum proteins for deter-
mination of glycemic control are: (1) control can
be measured consistently over a shorter time
frame because of the shorter half-life of serum
proteins (albumin 14 days), and (2) serum pro-
teins could be used in instances in which RBC
lifespans (and accuracy of GHbs) are altered.
Comparisons of glycated proteins (Fr and GPP)
to GHb in patients with diabetes mellitus without
concurrent renal diseases have been reported.37,38
The validity of the Fr assay and its use in
assessment of short-term glycemic control have
also been described.29,30
The DCCT data showed the following regres-
sion relationship between average BG andA1C15:
BG ϭ 30.9 (A1C) Ϫ 60.2
Based on these data, there was a trend for our
hemodialysis patients to have lower BG concen-
trations than the DCCT patients with normal
renal function for A1Cs greater than approxi-
mately 7.5%. In hemodialysis patients, an A1C
increase of 1% related to a 20 mg/dL increase in
BG. This is in contrast to the DCCT in which an
increase in A1C of 1% was related to a 30 mg/dL
increase in BG.15,16 Thus, each 1% increase in
A1C represents a 33% lower BG in dialysis
patients than those without renal insufficiency.
Although we did not directly compare measure-
ment of A1C by the Tina-Quant and HPLC as-
says with our specimens, we were justified in
comparing our results with the DCCT data be-
cause there is a direct correlation, similar line of
identity, and small amount of bias between the 2
methods.26,27
DISCUSSION
The results of the DCCT showed the benefit of
intensive insulin therapy and BG assessment;
34% reduction in retinopathy, 35% reduction in
microalbuminuria, and 60% reduction in neurop-
athy in primary prevention.16 A difference in
mean GHb level of only 2% resulted in this
marked reduction in complications.16 Although
these profound reductions in complications have
been shown with routine monitoring, estimates
in the literature report that only about 30% to
40% of diabetic patients routinely self-monitor
their blood sugars.34,35
GHb has been routinely used to assess glyce-
mic control of patients with diabetes mellitus. It
is formed slowly and nonenzymatically from
hemoglobin and glucose. It is recognized as an
excellent marker of glycemic control because its
rate of formation is proportional to the BG con-
centration over the average RBC lifespan (mean
120 days). The validity of A1C in monitoring the
glycemic control of patients with ESRD has been
Hemoglobin A1C is a posttranslational modifi-
cation of hemoglobin A. The hemoglobin A
subfractions were initially measured by ion-
exchange chromatography and gel electrophore-
sis, but are now also measured by immunoassay
methods (eg, Tina-Quant).27 GHb describes all

304
JOY ET AL
GHb species as quantified by affinity chromatog-
raphy, which detects structural differences in-
stead of charge selectivity (ion exchange chroma-
tography and gel electrophoresis). Interferences
with GHb assay methods are common and are
listed in Table 1.17,18 In addition, the results of the
various assays measuring GHb are not compared
with a single GHb standard, thus limiting the
potential to directly compare results across all
tests and laboratories. A recent review of the
Tina-Quant A1C immunoassay (standardized to
the HPLC method used in the DCCT trial) showed
no significant interference from bilirubin and
triglycerides up to levels of 64.5 mg/dL and
2,000 mg/dL, respectively.26,27 None of the pa-
tients evaluated in our study had bilirubin or
triglyceride concentrations in the assay interfer-
ence range. In addition, no interference from
hemoglobin F was shown for this method.27 Our
regression analysis indicated nearly identical cor-
relation coefficients when BG was plotted against
either A1C or GHb.
Like GHb, the Fr and GPP methods also have
assay interferences that are delineated in Table
1.17 Immunoassay methods for Fr have been
developed to counter some of the interferences.
There has been considerable debate regarding
whether the Fr assay should be corrected for total
protein or albumin concentration, especially if
albumin concentration is below the normal
range.33,39-41 The GPP test measures albumin and
all other plasma proteins; as such, it may be less
influenced by variations in 1 protein. Both the Fr
and GPP tests are indicative of glycemic control
over the previous 1 to 3 weeks, instead of the
previous 8 to 16 weeks (range of RBC lifespans
in hemodialysis) because the half-lives of the
plasma proteins are less than that of the RBC.
Because the current guidelines (DOQI) recom-
mend an adequate protein intake in hemodialysis
patients to maintain a normal albumin concentra-
tion, hypoalbuminemia may become less of a
significant issue in assessing glycemic control
with methods based on glycated proteins or Fr.42
Nevertheless, we sought to limit the effect of
albumin concentration by eliminating patients
with albumin concentrations of less than 3.0
g/dL. In addition, we corrected our patient’s
albumin concentrations to the normal range when
graphically analyzing the correlations between
BG and Fr by using the method of Howey et al.33
The normalization of these concentrations had
no effect on these correlations. As suggested in
our results, additional patients to the data set may
improve the correlation between Fr and BG. The
lower reference ranges for all tests of glycemic
control were higher in our ESRD patients than
reported in patients without renal disease. In
addition, for Fr and GPP, a higher upper refer-
ence limit was shown in patients with ESRD.
Testing that measures glycated proteins ap-
pears to have less assay interference from car-
bamylated proteins when the thiobarbituric acid
method is used. To avoid assay interference with
our measurements of GPP, we used affinity HPLC
and a thiobarbituric acid method to measure GPP
and Fr, respectively. Avoidance of ion-exchange
and gel electrophoresis methods in measuring
glycated proteins is suggested in patients receiv-
ing hemodialysis. Although we sought to elimi-
nate interferences that might alter the interpreta-
tion of our results, the measurement of GPP did
not appear to correlate well with BG in our
ESRD patients.
Several articles have recently evaluated vari-
ous measures of glycemic control in patients
with chronic renal failure, though none ad-
dressed the issue of accuracy of these meth-
ods.20-22 Increases in hemoglobin A1 (13%–22%)
have been shown in uremic patients by ion-
exchange chromatography,22 but assay methods
such as HPLC, colorimetry, or immunoassay
appear to produce less elevations in GHb mea-
surements owing to the uremic state or the dialy-
sis therapy.20,21 We evaluated A1C and GHb by
immunoassay and affinity HPLC methods, re-
spectively, to eliminate interference with our
measurements. It is necessary to note that when
clinically evaluating the A1C, one needs to con-
sider the assay methodology used and reported
reference range. For instance, though the normal
reference range for the A1C immunoassay we
used was 4.0% to 5.5%, a published boronate
affinity chromatography assay report shows a
much wider normal reference range of 4.5% to
7.0%.21,31 Our data revealed that measurement of
GHb and A1C correlated well with mean BG
concentrations. GPP correlated poorly with aver-
age BG measurements.
To assess the accuracy of the A1C test, we
compared the relationship of A1C measurements

GLYCEMIC CONTROL MEASURES IN HEMODIALYSIS
305
in diabetic patients on dialysis (as determined in
our study) with the results for that test in patients
with preserved renal function (as determined in
the DCCT trial). There are several limitations to
this approach: different assay methodologies were
used to assess A1C, and we were unable to obtain
95% confidence interval data from the DCCT
trial. In light of the first of these theoretical
limitations, a comparison of the A1C results from
the Tina Quant (our study) and HPLC (DCCT)
methods showed nearly identical measures of
A1C (with a bias of Ϫ0.09 to Ϫ0.1 only).26
Comparison of the regression lines between A1C
in patients with preserved renal function (DCCT)
versus ESRD (our study) revealed that A1C tends
to underestimate glycemic control in the latter
group. Thus, for any value of A1C above 7.5%,
the serum glucose measurements tend to be lower
in ESRD patients on hemodialysis than in pa-
tients with preserved renal function (as defined
by the DCCT). The difference between the 2
A1C-versus-BG curves tends to increase as A1C
increases. On the contrary, A1C overestimates
mean glucose concentrations at values near and
below the target of 6.5% for patients with ESRD.
Because large trials similar to the DCCT have
not been reported using GHb, we were unable to
infer the accuracy of this measurement in predict-
ing BG in our dialysis patients.
Because the Tina-Quant assay was docu-
mented to be free from interference caused by
carbamylated hemoglobin and other interference
factors were not present, the increased reported
A1C in our dialysis patients was most likely
caused by uremia. Indeed, the difference in glu-
cose concentrations can be substantial at higher
A1C levels. Thus, for example, an A1C of 7.5%
would correspond to a BG of 159 mg/dL in a
patient with ESRD and 172 mg/dL in a diabetic
patient without renal failure. At A1Cs of less than
6%, the corresponding A1C per level of BG is
lower for patients receiving hemodialysis, pos-
sibly reflecting an influence from other docu-
mented confounders. The differences in mea-
sured A1C we found between the DCCT data and
our hemodialysis patients is preliminary data.
The magnitude of this difference may be less
apparent with greater numbers of evaluable he-
modialysis patients. Also, the DCCT incorpo-
rated data from quarterly evaluations for 6.5
years in 1,441 patients, which would tend to
tighten the fit of the regression line.
For clinical purposes, the character of this
relationship is such that A1C remains a very
useful test to guide the therapy of patients on
hemodialysis. A hemodialysis patient with an
A1C near the recommended target of 7.0% would
have better glycemic control than a patient with-
out compromised renal function. On the other
hand, a very elevated A1C value (eg, Ͼ8.5%)
would still relay the message of poor glycemic
control to the patient and treating physician.
Also, for centers that report only GHb, it is
reasonable to feel confident in the correlation
with BG in patients with ESRD.
Since the length of the RBC lifespan is pre-
sumed to determine the accuracy of A1C and
GHb measurements because of exposure time to
BG concentrations, assessment of the lifespan
may be useful to evaluate these measures. The
life of the RBC in dialysis patients may be
altered as compared with the normal lifespan of
120 days in patients not receiving dialysis and
without renal insufficiency. One model suggests
a mean RBC lifespan of 68 days in hemodialysis
patients with an interindividual variability of
34%.25 RBCs with a shortened lifespan have a
shorter exposure to the ambient glucose concen-
trations in the blood. This may result in a reduced
measured A1C per level of glycemia when com-
pared with patients with normal RBC lifespans.
Our patients had average RBC lifespans of 53
days versus 120 days in patients without renal
disease.25 However, based on our regression anal-
ysis, a good correlation between RBC lifespan
and A1C failed to exist. The poor fit of the
erythrokinetic model in our patients (mean re-
sidual of plot of the regression line ϭ 1.22 Ϯ
.55) may reflect the fact that RBC lifespan has
less of an impact on measured A1C compared
with BG concentrations and/or the presence of
other concurrent factors. It may also be the result
of our relatively small sample size and narrow
range of RBC lifespans. This model was not
improved significantly with the addition of the
repeated measurements that were performed in a
small group of patients or by controlling for level
of BG. Based on our limited data, there is little
use in predicting RBC lifespan to predict accu-
racy of A1C or GHb measurements.

306
JOY ET AL
CONCLUSIONS
restriction and blood-pressure control on the progression of
chronic renal disease. N Engl J Med 330:877-884, 1994
10. Peterson JC, Adler S, Burkart JM, Greene T, Hebert
LA, Hunsicker LG, King AJ, Klahr S, Massry SG, Seifter
JL: Blood pressure control, proteinuria, and the progression
of renal disease. Ann Intern Med 123:754-762, 1995
11. Kshirsagar AV, Joy MS, Hogan SL, Falk RJ, Colin-
dres RE: Effects of angiotensin converting enzyme inhibi-
tors in diabetic and nondiabetic chronic renal disease: A
systematic overview of randomized placebo-controlled tri-
als. Am J Kidney Dis 35:695-707, 2000
12. Best JD, O’Neal DN: Diabetic dyslipidemia: Current
treatment recommendations. Drugs 59:1101-1111, 2000
13. Anonymous: Smoking and diabetes:American Diabe-
tes Association. Diabetes Care 23:93-94, 2000
14. The DCCT Research Group: Diabetes control and
complications trial (DCCT): Results of feasibility study.
Diabetes Care 10:1-19, 1987
15. Goldstein DE, Wiedmeyer HM, Little RR, Vargas V,
Nair SS, Reid J: Relationship between glycohemoglobin
(GHB) and mean blood glucose (MBG) in the diabetes
control and complications trial (DCCT). Diabetes 46:8A,
1997
16. The Diabetes Control and Complications Trial Re-
search Group: The effect of intensive treatment of diabetes
on the development and progression of long-term complica-
tions in insulin-dependent diabetes mellitus. N Engl J Med
329:977-986, 1993
Glycated hemoglobin (A1C or GHb) appears
to correlate most accurately with measured BG
in patients with ESRD receiving hemodialysis,
whereas Fr and GPP, as measures of GPP, appear
to correlate poorly with glycemic control. Our
study implies that there is an underestimation of
glycemic control by assessment of A1C in hemo-
dialysis patients, especially when A1C is greater
then 7.5%. If the benefits of tight glycemic
control as shown by the DCCT trial apply to the
population of patients on dialysis, a higher A1C
in hemodialysis patients is less worrisome than
in the overall population because this represents
a lower BG concentration. Hemodialysis patients
with good glycemic control, as defined byA1C of
less than 6.0%, appear to have higher levels of
BG than that published for the DCCT in patients
without coexisting renal disease, suggesting a
role of possible confounders in the immunoassay
method. The reference ranges and specificity
need to be considered when interpreting the
results of measurements of glycemic control in
ESRD patients.
17. Goldstein DE, Little RR, Lorenz RA, Malone JI,
Nathan D, Peterson CM: Tests of glycemia in diabetes.
Diabetes Care 18:896-909, 1995
REFERENCES
18. Service JF: Correlation between glycemia and gly-
cated hemoglobin. Compr Ther 16:33-40, 1990
19. Tzamaloukas AH: Interpreting glycated hemoglobin
in diabetic patients on peritoneal dialysis. Adv Perit Dial
12:171-175, 1996
20. Sabater J, Quereda C, Herrera I, Pascual J, Villafuela
JJ, Ortuno J: Nonenzymatic glycosylation of hemoglobin
and total plasmatic proteins in end-stage renal disease. Am J
Nephrol 11:37-43, 1991
21. Morgan L, Marenah BC, Jeffcoate WJ, Morgan AG:
Glycated proteins as indices of glycaemic control in diabetic
patients with chronic renal Failure. Diabet Med 13:514-519,
1996
1. US Renal Data System: USRDS 2000 Annual Data
Report. Bethesda, MD, National Institutes of Health, Na-
tional Institute of Diabetes and Digestive and Kidney Dis-
eases, 2000
2. Tuttle KR, Stein JH, DeFronzo RA: The natural his-
tory of diabetic nephropathy. Semin Nephrol 10:184-193,
1990
3. National Diabetes Data Group: Diabetes in America:
Diabetes data compiled 1984 (NIH publication no. 85-
1468). Bethesda, MD: National Institutes of Health, 1985
4. Deckert T, Poulsen JE, Larsen M: Prognosis of diabe-
tes with diabetes onset before the age of thirty-one. Diabeto-
logia 14:363-377, 1978
22. Smith WG, Holden M, Benton M, Braun CB: Gly-
cated and carbamylated haemoglobin in uraemia. Nephrol
Dial Transplant 4:96-100, 1989
5. Engerman R, Bloodworth JM Jr, Nelson S: Relation-
ship of microvascular disease in diabetes to metabolic con-
trol. Diabetes 26:760-769, 1977
23. Panzer S, Kronik G, Lechner K, Bettelheim P, Neu-
mann E, Dudczak R: Glycated hemoglobins (Ghb): An
index of red cell survival. Blood 59:1348-1350, 1982
24. NKF-DOQI clinical practice guidelines for hemodi-
alysis adequacy. National Kidney Foundation-Dialysis Out-
comes Quality Initiative. Am J Kidney Dis 30:S1-S66, 1997
(suppl 2)
25. Uehlinger DE, Gotch FA, Sheiner LB:Apharmacody-
namic model of erythropoietin therapy for uremic anemia.
Clin Pharmacol Ther 51:76-89, 1992
6. Maki DD, Ma JZ, Louis TA, Kasiske BL: Long term
effects of antihypertensive agents on proteinuria and renal
function. Arch Intern Med 155:1073-1080, 1995
7. Ravid M, Savin H, Jutrin I, Bental T, Katz B, Lishner
M: Long term stabilizing effect of angiotensin converting
enzyme inhibition on plasma creatinine and on proteinuria in
normotensive type II diabetic patients. Ann Intern Med
118:577-581, 1993
8. Parving HH, Rossing P, Monnel E, Smidt UA: Angio-
tensin converting enzyme inhibition in diabetic nephropa-
thy: Ten years’ experience. Am J Kidney Dis 26:99-107,
1995
26. Product literature For Tina-Quant® HbA1C immuno-
assay. Indianapolis, IN, Boehringer Mannheim
9. Klahr S, Levey AS, Beck GJ, Caggiula AW, Hunsicker
L, Kusek JW, Stricker G: The effects of dietary protein
27. Shuler C, Goldstein D, Little R, Wiedmeyer HM,
Wilke A, Cook J, Frye R, Vellasco S, Hoffman K, Miller R:

GLYCEMIC CONTROL MEASURES IN HEMODIALYSIS
307
Evaluation of a Tina-Quant HbA1c immunoassay on BM/
Hitachi analyzers. Boehringer Mannheim Corporation Publi-
cation. Indianapolis, IN, Boehringer Mannheim
36. LevinA, Foley RN: Cardiovascular disease in chronic
renal insufficiency. Am J Kidney Dis 36:S24-S30, 2000
(suppl 3)
28. Cefalu WT, Wang ZQ, Bell-Farrow A, Kinger FD,
Izlan C: Glycohemoglobin measured by automated affinity
HPLC correlates with both short-term and long-term anteced-
ent glycemia. Clin Chem 40:1317-1321, 1994
37. Smart LM, Howie AF, Young RJ, Walker SW, Clarke
BF, Smith AF: Comparison of fructosamine with glycated
hemoglobin and plasma proteins as measures of glycemic
control. Diabetes Care 11:433-436, 1988
29. Cefalu WT, Bell-Farrow AD, Petty M, Islar C, Smith
JA: Clinical validation of second-generation fructosamine
assay. Clin Chem 37:1252-1256, 1991
30. Cefalu WT, Parker TB, Johnson CR: Validity of
serum fructosamine as index of short-term glycemic control
in diabetic outpatients. Diabetes Care 11:662-664, 1988
31. Manual of Pathology and Laboratory Medicine Clini-
cal Services. University Printing, 1997
32. Schleicher ED, Olgemoller B, Wiedenmann E, Ger-
bitz KD: Specific glycation of albumin depends on its
half-life. Clin Chem 39:625-628, 1993
33. Howey JEA, Browning MCK, Fraser CG: Assay of
serum fructosamine that minimises standardisation and
manual problems: Use to assess components of biological
variation. Clin Chem 33:269-272, 1987
38. Koskinen P, Erkkola R, Viikari J, Mattila K, Irjala K:
Blood glycated haemoglobin, serum fructosamine, serum
glycated albumin and serum glycated total protein as mea-
sures of glycaemia in diabetes mellitus. Scand J Clin Lab
Invest 52:863-869, 1992
39. Baker JR, O’Connor JP, Metcalf PA, Lawson MR,
Johnson RN: Clinical usefulness of estimation of serum
fructosamine concentration as a screening test for diabetes
mellitus. BMJ 287:863-867, 1983
40. Johnson RN, Metcalf PF, Baker JR: Relationship
between albumin and fructosamine concentration in diabetic
and nondiabetic sera. Clin Chim Acta 164:151-162, 1987
41. Van dieijen-Visser MP, Seynaeve C, Brombacher PJ:
Influence of variations in albumin or total protein concentra-
tion on serum fructosamine concentration. Clin Chem 32:
1610, 1986 (letter)
42. NKF-DOQI clinical practice guidelines for nutrition
in chronic renal Failure. National Kidney Foundation-
Dialysis Outcomes Quality Initiative. Am J Kidney Dis
35:S1-S103, 2000 (suppl 2)
34. Harris MI: Testing of blood glucose by office-based
physicians in the US. Diabetes Care 13:419-426, 1990
35. Harris MI, Cowie CC, Howie LF: Self-monitoring of
blood glucose by adults with diabetes in the United States
population. Diabetes Care 16:1116-1123, 1993